Machine Learning Engineer

На данный момент Machine Learning Engineer.

Мои компетенции и опыт

Experienced Machine Learning Engineer with over 5 years of expertise in Computer Vision, adept at developing cutting-edge deep learning solutions across healthcare, video surveillance, and manufacturing sectors. Possesses deep knowledge in enhancing model performance, and driving successful project deployments. Developed advanced ML systems and significantly improved IoU. Skilled in model training, R&D projects and data analysis with a strong focus on scalability and reproducibility. Adaptable and proficient in the latest AI and ML technologies, making significant contributions to various stages of the project lifecycle.

• Integrated an ML solution for the healthcare industry using Self-supervised Learning and Vision Transformers (ViT), boosting the quality of pathology recognition for 3 medical tasks by 5% and reducing the models' training time by 10x in 1 year.
• Engineered a fight detection system based on pose estimation and optimized AlphaPose using ONNX and TensorRT, accelerating the model's inference by 5x in 3 months.
• Prepared data and trained a deep learning model for teeth 3D segmentation, achieving 94% accuracy on the test dataset.
• Explored state-of-the-art approaches as an R&D project to anomaly detection in video and trained a model based on an unsupervised learning approach to detect anomalies within 2 months.
• Developed an inspection system for the restoration of buildings based on object detection and segmentation, improving IoU from 0.6 to нужен доступ к резюме
• Improved model quality for knee arthrosis recognition by 5% by integrating custom layers into the model architecture in 2 weeks.
• Enhanced visualization of CT images using HDR, which allowed the doctor to view pathologies requiring different window settings in one window.
• Developed a computer vision system for the steel industry using OpenCV and Pytorch, resulting in an MVP service for the customer in 4 months.
• Created an end-to-end anti-fraud ML service for cinemas, detecting seat violations and ticketless entry, and successfully transitioned the project from concept to real-customer product deployment with 30+ connected halls.